High Performance LDPC Decoder Using a Simplified Robust Algorithm in the Categorized Tanner Graph

Abstract:
This paper presents an LDPC decoder design method for avoiding a high error floor caused by the approximation in the Max-Log domain known as the min-sum algorithm in the categorized tanner graph. This decoder, which doesn’t need any knowledge of multi-path channels, has the benefit of robustness over level mismatch between demodulator and decoder. To overcome error floor, we apply some individual coefficients to LLR values in every category. Simulation results in AWGN and multi-path fading channels show that even though categorized low profile and approximations are used, good performance is achieved, which cannot be achieved by ordinary single coefficient methods (e.g., normalized BP algorithm). Unlike the recently developed simplified algorithms which use the noise variance of the channel estimate, simulation results show the proposed multi-coefficient method which doesn’t use the noise variance is robust over the varying fading conditions.